def test_auto_backend():
"""Check the backend detection logic."""
original_env_backend = os.environ.get('CLUSTERLIB_BACKEND', None)
if original_env_backend is not None:
del os.environ['CLUSTERLIB_BACKEND']
try:
# Check detection when no environment variable is set.
if _which('scontrol'):
# SLURM should be detected
assert_equal(_get_backend('auto'), 'slurm')
elif _which('qmod'):
# SGE should be detected
assert_equal(_get_backend('auto'), 'sge')
else:
# No backend can be detected
assert_raises(RuntimeError, _get_backend, 'auto')
# Check the use of the environment variable
os.environ['CLUSTERLIB_BACKEND'] = 'slurm'
assert_equal(_get_backend('auto'), 'slurm')
os.environ['CLUSTERLIB_BACKEND'] = 'sge'
assert_equal(_get_backend('auto'), 'sge')
finally:
# Restore the previous environment
if original_env_backend is None:
del os.environ['CLUSTERLIB_BACKEND']
else:
os.environ['CLUSTERLIB_BACKEND'] = original_env_backend
def test_auto_backend():
"""Check the backend detection logic."""
original_env_backend = os.environ.get('CLUSTERLIB_BACKEND', None)
if original_env_backend is not None:
del os.environ['CLUSTERLIB_BACKEND']
try:
# Check detection when no environment variable is set.
if _which('scontrol'):
# SLURM should be detected
assert_equal(_get_backend('auto'), 'slurm')
elif _which('qmod'):
# SGE should be detected
assert_equal(_get_backend('auto'), 'sge')
else:
# No backend can be detected
assert_raises(RuntimeError, _get_backend, 'auto')
# Check the use of the environment variable
os.environ['CLUSTERLIB_BACKEND'] = 'slurm'
assert_equal(_get_backend('auto'), 'slurm')
os.environ['CLUSTERLIB_BACKEND'] = 'sge'
assert_equal(_get_backend('auto'), 'sge')
finally:
# Restore the previous environment
if original_env_backend is None:
del os.environ['CLUSTERLIB_BACKEND']
else:
os.environ['CLUSTERLIB_BACKEND'] = original_env_backend
def check_job_name_queued_or_running(job_name):
with TemporaryDirectory() as temp_folder:
user = getuser()
# Launch job
command = submit(job_command="sleep 600",
job_name=job_name,
time="700",
memory=500,
log_directory=temp_folder)
job_id = _check_job_id(command)
# Assert that the job has been launched
try:
running_jobs = queued_or_running_jobs(user=user)
assert_in(job_name, running_jobs)
finally:
# Make sure to clean up even if there is a failure
if _get_backend() == "slurm":
subprocess.call(["scancel", job_id])
elif _get_backend() == "sge":
subprocess.call(["qdel", job_id])
else:
raise NotImplementedError("backend not implemented")
def test_fixed_backend():
"""Check that it is possible to fix explicit backends (when valid)."""
# Supported backends
assert_equal(_get_backend('slurm'), 'slurm')
assert_equal(_get_backend('sge'), 'sge')
# Unsupported backend
assert_raises(ValueError, _get_backend, 'hadoop')
def test_fixed_backend():
"""Check that it is possible to fix explicit backends (when valid)."""
# Supported backends
assert_equal(_get_backend('slurm'), 'slurm')
assert_equal(_get_backend('sge'), 'sge')
# Unsupported backend
assert_raises(ValueError, _get_backend, 'hadoop')
def check_job_name_queued_or_running(job_name):
with TemporaryDirectory() as temp_folder:
user = getuser()
# Launch job
command = submit(job_command="sleep 600", job_name=job_name,
time="700", memory=500, log_directory=temp_folder)
job_id = _check_job_id(command)
# Assert that the job has been launched
try:
running_jobs = queued_or_running_jobs(user=user)
assert_in(job_name, running_jobs)
finally:
# Make sure to clean up even if there is a failure
if _get_backend() == "slurm":
subprocess.call(["scancel", job_id])
elif _get_backend() == "sge":
subprocess.call(["qdel", job_id])
else:
raise NotImplementedError("backend not implemented")
def test_log_output(n_trials=30):
"""Test that log output is uniform accross scheduler."""
with TemporaryDirectory() as temp_folder:
user = getuser()
job_completed = False
# Launch a sleepy SGE job
job_name = 'ok_job'
command = submit(job_command="echo ok",
job_name=job_name,
time="700",
memory=500,
log_directory=temp_folder)
job_id = _check_job_id(command)
try:
for _ in range(n_trials):
if job_name not in queued_or_running_jobs(user=user):
# job has completed, let's check the output
job_completed = True
filename = "%s.%s.txt" % (job_name, job_id)
assert_equal(os.listdir(temp_folder), [filename])
with open(op.join(temp_folder, filename)) as fhandle:
assert_equal(fhandle.read().strip(), "ok")
break
else:
# Let's wait a bit before retrying
sleep(5)
finally:
# Make sure to clean up even if there is a failure
if not job_completed:
if _get_backend('auto') == 'slurm':
subprocess.call(["scancel", job_id])
else:
subprocess.call(["qdel", job_id])
raise AssertionError(
"job %s (%s) has not completed after 5min." %
(job_id, job_name))
def test_log_output(n_trials=30):
"""Test that log output is uniform accross scheduler."""
with TemporaryDirectory() as temp_folder:
user = getuser()
job_completed = False
# Launch a sleepy SGE job
job_name = 'ok_job'
command = submit(job_command="echo ok", job_name=job_name,
time="700", memory=500,
log_directory=temp_folder)
job_id = _check_job_id(command)
try:
for _ in range(n_trials):
if job_name not in queued_or_running_jobs(user=user):
# job has completed, let's check the output
job_completed = True
filename = "%s.%s.txt" % (job_name, job_id)
assert_equal(os.listdir(temp_folder), [filename])
with open(op.join(temp_folder, filename)) as fhandle:
assert_equal(fhandle.read().strip(), "ok")
break
else:
# Let's wait a bit before retrying
sleep(5)
finally:
# Make sure to clean up even if there is a failure
if not job_completed:
if _get_backend('auto') == 'slurm':
subprocess.call(["scancel", job_id])
else:
subprocess.call(["qdel", job_id])
raise AssertionError(
"job %s (%s) has not completed after 5min."
% (job_id, job_name))